Re: [Chicken-users] State machine using call/cc

[Joel Reymont]

> (define (next-state game)
>   (call/cc
>     (lambda (k)
>       (game-continuation-set! game k)
>       ((game-next-turn game) game) ) ) )

So in essence and if I understand this correctly, each state in the machine 
needs to save its continuation once so that it can be re-entered. Right?

What if there is some data saved in the machine (continuation?)  and
different states updated that. How would I ensure that they all got the
same continuation or will they get it automatically?

Lets assume that the data stored in the machine (game variables like a
list of players in the game) was the "context". I could do the state
machine by returning a pair of (data-to-be-sent . new-state-proc) from
the dispatcher but then I would always need to pass this context together
with the request/event into a state (proc) to get the next state.

The advantage of doing the same with continuations seems to be in that I
don't need to return the pair (data-to-be-sent . new-state-proc) from the
request dispatcher. I can just return the data-to-be-sent after saving
new-state-proc as a continuation. Then I would not need to pass the
context into  new-state-proc when calling it with the next event as i
would retrieve this context as part of the continuation before calling
new-state-proc.

Am I correct in my thinking?

>;;; this is probably called from C/C++:
>(define (process-request req)
>  (generate-response
>    (call/cc
>      (lambda (k)
>          (game-next-turn-set! (request->game req) k)
>          ((game-continuation (request->game req)) req) ) ) ) )

If I had multiple state machines then I would also need to save each one
as a continuation it seems, right?

    Joel

-- 
OpenPoker: The Linux of poker software
http://wagerlabs.com/forums